Railway vehicle and vehicle energy absorption structure crashworthiness collaborative optimization method and system

Abstract

The present application relates to the field of rail vehicle technology, and discloses a rail vehicle and vehicle energy-absorbing structure crashworthiness collaborative optimization method and system, comprising: obtaining impact characteristic data sets of the energy-absorbing structure under different structural parameter configurations, and establishing a structural level deep proxy model to predict the load characteristic curve and deformation mode image of the energy-absorbing structure; inputting the load characteristic curve of the energy-absorbing structure output by the structural level deep proxy model into a vehicle collision dynamics model, carrying out vehicle level collision simulation, and constructing a vehicle level deep proxy model to predict the vehicle collision response; based on the deformation mode image obtained by the structural level deep proxy model and the vehicle collision response obtained by the vehicle level deep proxy model, performing multi-objective rail vehicle / energy-absorbing structure crashworthiness collaborative constraint optimization to obtain an optimization scheme; the present application solves the problems of low calculation efficiency and poor reliability of the optimization results of the existing collision optimization method of the rail vehicle and the energy-absorbing structure.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of rail vehicle technology, in particular to a rail vehicle and vehicle energy-absorbing structure crashworthiness collaborative optimization method and system. BACKGROUND

[0002] The rail vehicle collision is a strong nonlinear and complex system dynamics response process, which includes not only the energy-absorbing structure absorbing impact kinetic energy through its own plastic deformation, but also the nonlinear evolution of the attitude of the vehicle structure under the action of multiple source loads. Therefore, comprehensively considering the deformation behavior of the energy-absorbing structure and the stability of the vehicle attitude evolution is an effective means to ensure the reliability of the impact behavior of the rail vehicle and the stability of the energy dissipation. At present, the traditional method based on finite element model simulation for scheme iterative optimization has the most core problem of difficult to meet the engineering requirements in terms of calculation efficiency when dealing with the optimization of the crashworthiness of the new energy structure of the rail vehicle. In order to improve the calculation and optimization efficiency, the prior method mainly simplifies the collision dynamics model to carry out optimization, uses the dynamics model to simplify the energy-absorbing structure entity into a collision mechanical property curve and inputs it into the dynamics model, and carries out optimization of the collision energy dissipation and impact attitude response of the rail vehicle based on the model. However, in this way, the collision response of the energy-absorbing structure is only represented by the mechanical property curve, the entity structure parameters corresponding to the mechanical property curve cannot be directly obtained, and further reverse calculation is required, and whether the real collision response of the structure has non-ideal conditions such as unstable deformation mode and severe load fluctuation is also completely unknown. It can be seen that the existing collision optimization method of the rail vehicle and the energy-absorbing structure has the problems of low calculation efficiency and poor reliability of the optimization results. SUMMARY

[0003] The present application provides a rail vehicle and vehicle energy-absorbing structure crashworthiness collaborative optimization method and system to solve the problem of low calculation efficiency and poor reliability of the optimization results of the existing collision optimization method of the rail vehicle and the energy-absorbing structure.

[0004] In order to achieve the above-mentioned purpose, the present application realizes the technical scheme as follows:

[0005] In the first aspect, the present application provides a rail vehicle and vehicle energy-absorbing structure crashworthiness collaborative optimization method, comprising:

[0006] Through impact test and finite element simulation of the energy-absorbing structure, impact characteristic data sets of the energy-absorbing structure under different structure parameter configurations are obtained, and a structure level deep proxy model is established based on the impact characteristic data sets of the energy-absorbing structure to predict the load characteristic curve and the deformation mode image of the energy-absorbing structure;

[0007] The load characteristic curve of the energy-absorbing structure output by the structure-level deep agent model is input into a vehicle collision dynamics model, vehicle-level collision simulation is carried out, and a vehicle-level deep agent model is constructed to predict vehicle collision response;

[0008] Based on the deformation mode image obtained by the structure-level deep agent model and the vehicle collision response obtained by the vehicle-level deep agent model, multi-objective railway vehicle / energy-absorbing structure crashworthiness collaborative constraint optimization is carried out, and a railway vehicle crashworthiness optimization scheme with stable energy-absorbing structure deformation and controllable vehicle posture evolution is obtained.

[0009] Optionally, the step of establishing the structure-level deep agent model comprises:

[0010] According to the configuration type of the energy-absorbing structure of the railway vehicle, the key parameters affecting the collision energy-absorbing response of the structure are determined, the structure impact test and corresponding numerical simulation under different parameter configurations are carried out, and the simulation model is calibrated;

[0011] Based on the calibrated numerical simulation model of the test, an impact characteristic data set of the energy-absorbing structure is generated, and is divided into a training set, a verification set and a test set;

[0012] Based on a deep learning framework and the constructed impact characteristic data set of the energy-absorbing structure, a structure-level deep agent model is built, and the training set is used to train the structure-level deep agent model, and the verification set and the test set are used to verify and test the trained structure-level deep agent model;

[0013] The structure-level deep agent model comprises an impact load curve network and a deformation mode image network, the impact load curve network is used to predict the load characteristic curve of the energy-absorbing structure, and the deformation mode image network is used to predict the deformation mode image.

[0014] Optionally, the construction of the impact load curve network comprises:

[0015] An auto-encoding network with a stacked long short-term memory layer as the main architecture is constructed, and hidden variables of the load characteristic curve are extracted;

[0016] A multilayer perceptron vector prediction network is constructed, the corresponding extracted load characteristic curve hidden variables are predicted from the structure parameters, and the decoder in the auto-encoding network is decoded to output the load curve.

[0017] Optionally, the construction of the deformation mode image network comprises:

[0018] An auto-encoding network with a stacked convolutional layer as the main architecture is constructed, and hidden variables of the deformation mode image are extracted;

[0019] A multi-layer perceptron vector prediction network is constructed to predict the corresponding extracted deformation mode image hidden variable from the structural parameter, and the deformation mode image is decoded by a decoder in the auto-encoding network to output the deformation mode image.

[0020] Optionally, the step of constructing the vehicle-level deep agent model comprises:

[0021] The energy absorption structure load characteristic curve output by the structural-level deep agent model is input into a vehicle crash dynamics model, vehicle impact tests and corresponding numerical simulations under different parameter configurations are carried out, and the simulation model is calibrated;

[0022] Based on the calibrated numerical simulation model, a vehicle crash response dataset is generated, and is divided into a training set, a validation set and a test set;

[0023] Based on the deep learning framework and the constructed vehicle crash response dataset, a vehicle-level deep agent model is built, and the energy absorption structure parameters are used as inputs of the vehicle-level deep agent model to predict the vehicle crash response.

[0024] Optionally, the step of multi-objective rail vehicle / energy absorption structure crashworthiness collaborative constraint optimization comprises:

[0025] The design variable is the key parameter of the energy absorption structure, and the optimization objective is the maximum average deceleration and total energy absorption during the vehicle crash process;

[0026] The constraint conditions include energy absorption structure impact characteristic constraints and vehicle crash response constraints, the energy absorption structure impact characteristic constraints include deformation mode stability constraints and load smoothness constraints, and the vehicle crash response constraints include maximum wheelset vertical lift constraints;

[0027] Non-dominated sorting genetic algorithm, particle swarm optimization algorithm or differential evolution algorithm are used for multi-objective constraint optimization to obtain a Pareto solution set, and the shortest distance method or entropy weight method is used to calculate the optimal feasible solution.

[0028] Optionally, the deformation mode stability constraint in the energy absorption structure impact characteristic constraint is defined as:

[0029] If the deformation mode is stable, return the constraint value K=0, if the deformation mode is unstable, return the constraint value K=1.

[0030] Optionally, the load smoothness constraint in the energy absorption structure impact characteristic constraint is defined as:

[0031] ;

[0032] In the formula, S 2 is the fluctuation evaluation index of the load characteristic curve, θa load stability constraint value of the energy-absorbing structure, N a number of sampling points of the load curve, F i a load value at an arbitrary sampling time, i a sampling time, an average value of the load curve;

[0033] If S 2 ≤ θ, return a constraint value 0, and if S 2 > θ, return a constraint value 1, wherein θ is a preset constraint value.

[0034] Optionally, the non-dominated sorting genetic algorithm, the particle swarm algorithm or the differential evolution algorithm is used for multi-objective constraint optimization to obtain a Pareto solution set, including:

[0035] An energy-absorbing structure crashworthiness collaborative optimization mathematical model of the railway vehicle is constructed according to the design variables, the optimization objectives and the constraint conditions, and the mathematical model satisfies the following relationship:

[0036]

[0037] In the formula, X is a parameter of the energy-absorbing structure, X L and X U are upper and lower boundaries thereof respectively, θ a load stability constraint value of the energy-absorbing structure, K is a mode stability constraint of the energy-absorbing structure, and MVL is a wheelset lifting amount constraint of the vehicle;

[0038] The non-dominated sorting genetic algorithm, the particle swarm algorithm or the differential evolution algorithm is used to solve the constructed energy-absorbing structure crashworthiness collaborative optimization mathematical model to obtain a Pareto solution set.

[0039] The maximum wheelset vertical lifting amount constraint in the vehicle collision response constraint is defined as:

[0040] According to the railway vehicle crashworthiness design standard, the maximum wheelset vertical lifting amount in the collision process should be ≤21mm.

[0041] In a second aspect, the embodiments of the present application provide a railway vehicle and energy-absorbing structure crashworthiness collaborative optimization system, including a processor, a memory and the like.

[0042] The memory is used to store a computer program.

[0043] The processor is used to execute the program stored in the memory to realize the method steps in the first aspect.

[0044] Advantages:

[0045] ​The rail vehicle and vehicle energy absorption structure crashworthiness collaborative optimization method provided by the application first obtains a crash mechanics response curve and a deformation mode image data set under different structural parameter configurations through impact test and finite element simulation of the energy absorption structure, and establishes a deep proxy model at the structural level based on the data set, so as to realize prediction of the load characteristic curve and the deformation mode image of the energy absorption structure parameter configuration; the obtained load curve is input into a vehicle crash dynamics model, vehicle-level crash simulation is carried out, and a deep proxy model at the vehicle level is constructed, so as to realize prediction of the vehicle crash response of the energy absorption structure parameter. Finally, the mode image obtained by the deep proxy model at the structural level and the vehicle crash response obtained by the vehicle-level proxy model are used as optimization objectives and constraint conditions for vehicle / energy absorption structure crashworthiness collaborative optimization, so as to obtain a rail vehicle crashworthiness optimization scheme with stable deformation of the energy absorption structure and controllable vehicle posture evolution, and the reliability of the optimization result is greatly improved.

[0046] By using the proposed multi-level deep proxy modeling method, the energy absorption structure parameter can predict the structural impact characteristic curve and the deformation mode image, and the vehicle crash response can also be quickly predicted. Meanwhile, by using the proposed collaborative optimization method, the optimization result can meet the requirements of the vehicle crash response index, while ensuring that the deformation mode of the energy absorption structure is stable and the load fluctuation is smooth, and the reliability of the optimization result is greatly improved. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 A flowchart of the rail vehicle and vehicle energy absorption structure crashworthiness collaborative optimization method of the preferred embodiment of the application;

[0048] Figure 2 A schematic diagram of the network architecture of the deep proxy model at the structural level of the preferred embodiment of the application;

[0049] Figure 3 A schematic diagram of the load curve prediction network architecture of the deep proxy model at the structural level of the preferred embodiment of the application;

[0050] Figure 4 A schematic diagram of the mode image prediction network architecture of the deep proxy model at the structural level of the preferred embodiment of the application;

[0051] Figure 5 A configuration schematic diagram of an energy absorption structure at the end of a certain subway head car of the preferred embodiment of the application;

[0052] Figure 6 A schematic diagram of key structural parameters of each energy absorption structure of the preferred embodiment of the application;

[0053] Figure 7 A calibration result schematic diagram of an energy absorption structure model of the preferred embodiment of the application;

[0054] Figure 8 The force-displacement curve prediction effect schematic diagram of the energy-absorbing structure of the preferred embodiment of the present application;

[0055] Figure 9 The deformation mode image prediction effect schematic diagram of the energy-absorbing structure of the preferred embodiment of the present application;

[0056] Figure 10 The vehicle-level collision simulation model benchmarking result schematic diagram of the preferred embodiment of the present application;

[0057] Figure 11 The vehicle collision response prediction effect schematic diagram of the preferred embodiment of the present application;

[0058] Figure 12 The optimization result comparison diagram of the preferred embodiment of the present application. DETAILED DESCRIPTION

[0059] The technical solutions of the present application will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0060] Unless otherwise defined, the technical terms or scientific terms used in the present application shall have the meanings commonly understood by those skilled in the art to which the present application belongs. The terms "first", "second", and similar terms used in the present application do not represent any order, number, or importance, but are only used to distinguish different components. Similarly, the terms "one" or "a" and similar terms do not represent a quantity limitation, but represent the existence of at least one. The terms "connected" or "connected" and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The terms "up", "down", "left", "right", and the like only represent relative positional relationships, which change accordingly when the absolute positions of the described objects change.

[0061] Embodiment 1

[0062] Please refer to Figure 1 The present application embodiment provides a rail vehicle and vehicle energy-absorbing structure crashworthiness collaborative optimization method, comprising:

[0063] Through impact test and finite element simulation of the energy-absorbing structure, impact characteristic data sets of the energy-absorbing structure under different structural parameter configurations are obtained, and a structural level deep proxy model is established based on the impact characteristic data sets of the energy-absorbing structure to predict the load characteristic curve and deformation mode image of the energy-absorbing structure;

[0064] The energy absorption structure load characteristic curve output by the structure level deep agent model is input into a vehicle collision dynamics model, vehicle level collision simulation is carried out, and a vehicle level deep agent model is constructed to predict vehicle collision response;

[0065] Multi-objective railway vehicle / energy absorption structure crashworthiness collaborative constraint optimization is carried out based on the deformation mode image obtained by the structure level deep agent model and the vehicle collision response obtained by the vehicle level deep agent model, and an optimized scheme of railway vehicle crashworthiness is obtained, in which the deformation of the energy absorption structure is stable and the evolution of the vehicle posture is controllable.

[0066] In the above embodiment, the above method specifically comprises the following steps:

[0067] 1. Constructing a structure level deep agent model

[0068] ① According to the configuration type of the railway vehicle energy absorption structure, the key parameters (including the coupler device, anti-climbing energy absorption device and main energy absorption device) affecting the collision energy absorption response of the structure are determined, structure impact tests and corresponding numerical simulations under different parameter configurations are carried out, and the simulation model is calibrated:

[0069] ② Based on the calibrated numerical simulation model of the test, a test design is carried out, an energy absorption structure impact characteristic data set is generated, and the data set is randomly divided into a training set, a verification set and a test set according to an 8:1:1 ratio, wherein the training set is used for training the following prediction model, the verification set is used for adjusting the training parameters during training, and the test set is used for verifying the training effect of the model.

[0070] ③ Based on the deep learning framework and the data set constructed in the foregoing steps, an energy absorption structure multi-dimensional impact characteristic prediction model is built, and the model architecture is as shown in Figure 2 , (in the figure, part of the material is taken as an example of the energy absorption structure in the embodiment, and the same applies below), which is mainly divided into the following two parts:

[0071] ① Impact load curve network (such as Figure 3 ): first, a self-encoding network with stacked long short-term memory layers as the main architecture is constructed to extract load characteristic curve hidden variables ; a multi-layer perceptron vector prediction network is constructed to predict the corresponding hidden variables C from the structure parameters; then the decoder in the self-encoding network is decoded to output the obtained load curve;

[0072] ② Deformation mode image network (such as Figure 4 ): first, a self-encoding network with stacked convolutional layers as the main architecture is constructed to extract deformation mode image hidden variables ; a multi-layer perceptron vector prediction network is constructed to predict the corresponding hidden variables M from the structure parameters; then the decoder in the self-encoding network is decoded to output the obtained deformation mode image;

[0073] 2. Constructing a vehicle-level deep surrogate model

[0074] ① The energy-absorbing structure load characteristic curve output by the structural-level deep surrogate model is input into the vehicle crash dynamics model as a simplified representation of the energy-absorbing structure. Vehicle impact tests and corresponding numerical simulations are carried out under different parameter configurations, and the simulation model is calibrated;

[0075] ② Based on the calibrated numerical simulation model, a test design is carried out to generate a vehicle crash response dataset, which is randomly divided into training, validation, and test sets in a ratio of 8:1:1. The training set is used for training the prediction model, the validation set is used for adjusting the training parameters during training, and the test set is used to verify the training effect of the model;

[0076] ③ A vehicle crash response multilayer perception prediction model is built to predict vehicle crash response from energy-absorbing structure parameters;

[0077] 3. Multi-objective rail vehicle / energy-absorbing structure crashworthiness collaborative constraint optimization

[0078] ① Design variables: The design variables are the key parameters of the aforementioned energy-absorbing structures (coupler device, anti-climbing energy-absorbing device, and main energy-absorbing device);

[0079] ② Optimization objectives: The optimization objectives are the maximum average deceleration (MAD) and total energy absorption (TEA) during vehicle collision;

[0080] ③ Constraint conditions: The constraint conditions are divided into two parts, corresponding to the impact characteristics of the energy-absorbing structure and the vehicle crash response constraints.

[0081] (1) Energy-absorbing structure impact characteristic constraints: For energy-absorbing structures, the stability of the deformation mode and load history during impact is an important constraint condition. Based on the mode image prediction results, define the mode stability constraint, and according to the differences in energy-absorbing structure types, the stability criterion varies for different structures, such as mode stability, return constraint value K = 0, and mode instability, return constraint value K = 1; based on the load curve prediction results, define the load smoothness constraint value S 2 (As formula 1)

[0082] ;

[0083] Where S 2 is the fluctuation evaluation index of the load characteristic curve; θ is the preset constraint value of the index; N represents the number of sampling points of the load curve; F i represents the load value at any sampling time; The average value of the load curve is returned as the constraint value 0, and vice versa as the constraint value 1. , the constraint value 0 is returned, and vice versa, the constraint value 1 is returned;

[0084] (2) Vehicle collision response constraint: for rail vehicles, the stability of the attitude evolution in the impact process is an important constraint condition. According to the rail vehicle crashworthiness design standard (such as TB / T 3500, EN 15227), the maximum wheelset vertical lift (MVL) in the collision process should be ≤21mm;

[0085] ④ Multi-objective constraint optimization:

[0086] According to the definition of the design variables, optimization objectives and constraint conditions described above, a mathematical model of the rail vehicle / energy-absorbing structure crashworthiness collaborative optimization is established (such as formula 2), and general algorithms such as non-dominated sorting genetic algorithm (NSGA), particle swarm optimization algorithm (PSO), differential evolution algorithm (DE) are used to solve it, and the Pareto solution set is obtained;

[0087] ;

[0088] Where X represents the energy-absorbing structure parameters; X L and X U represent the upper and lower boundaries thereof, respectively; θ represents the energy-absorbing structure load stability constraint; K represents the energy-absorbing structure mode stability constraint; MVL represents the vehicle wheelset lift constraint.

[0089] After obtaining the multi-objective Pareto solution set, the optimal feasible solution is calculated by using general decision-making methods such as the shortest distance method and the entropy value weighting method, that is, the rail vehicle crashworthiness design scheme with controllable vehicle collision attitude and stable energy-absorbing structure deformation can be obtained.

[0090] Example 2

[0091] In this embodiment, the rail vehicle and the energy-absorbing structure crashworthiness of a certain subway head car are taken as an example for collaborative optimization, and it is assumed that the end of the certain subway head car is equipped with one set of coupler device, two anti-climbing energy-absorbing devices and two main energy-absorbing devices, which are shown in Figure 5 In the head car crashworthiness optimization problem, according to the technical solution, the following steps are mainly included:

[0092] 1. Build a structure level depth proxy model

[0093] ① First, define the key structure parameters (such as Figure 6 ) of each energy-absorbing structure, which include the coupler device: including wall thickness, bulging angle, initial diameter and bulging diameter, a total of 4 parameters; Anti-climbing energy-absorbing device: including honeycomb strength, honeycomb width and honeycomb height, a total of 3 parameters; Main energy-absorbing device: including honeycomb strength (4), folding pipe wall thickness, folding height and initial gap, a total of 7 parameters.

[0094] Car impact tests of the three energy absorption structures were carried out, and finite element model simulation was carried out. The calibration results of the simulation calculation model of each structure are as shown in Figure 7 .

[0095] ②Based on the calibrated calculation model, three kinds of energy absorption structure impact characteristic data sets were constructed respectively. Considering the complexity of the structure and the number of variables, the data set size of the car coupler device is 400, the data set size of the anti-climbing energy absorption device is 300, and the data set size of the main energy absorption device is 500. When constructing each data set, the change range of the structure parameters is shown in Table 1.

[0096] Table 1 Parameter change range of impact characteristic data set of each energy absorption structure

[0097]

[0098] ③The deep proxy model of the three kinds of energy absorption structures was constructed and trained respectively, and the force-displacement curve and deformation mode image were predicted from the structure parameters. The prediction effects are shown in Figure 8 and Figure 9 .

[0099] 2. Construction of vehicle level deep proxy model

[0100] ①The load characteristic curves of the three kinds of energy absorption structures output by the structure level deep proxy model were input into the vehicle crash dynamics model as a simplified representation of the energy absorption structure. Vehicle impact test and corresponding numerical simulation were carried out, and the simulation model was calibrated. The model calibration results are shown in Figure 10 .

[0101] ②Based on the calibrated calculation model, a vehicle crash response data set was constructed, in which the input was the key parameters of each energy absorption structure defined above, and the output was the maximum average deceleration MAD, total energy absorption TEA and maximum vertical lift MVL. The data set size was 500.

[0102] ③A vehicle crash response multilayer perception machine prediction model was built to predict vehicle crash response from energy absorption structure parameters. The prediction effects of each crash response are shown in Figure 11 .

[0103] 3. Multi-objective rail vehicle / energy absorption structure crashworthiness collaborative constraint optimization

[0104] According to the optimization process defined in the technical solution, the design variables, optimization objectives and constraint conditions of the optimization problem are determined, and the optimization mathematical model is clarified. Non-dominated sorting genetic algorithm (NSGA) is used for solution, and Pareto solution set is obtained. In order to illustrate the high reliability of the technical solution, the optimization results without introducing the constraint condition of the structure impact characteristic are compared (as shown in Figure 12). As can be seen from the figure, the optimization solution obtained without introducing the constraint of structural impact characteristics can meet the optimization effect of vehicle collision response, but there is an unstable deformation of the energy absorption structure, and by using the technical solution, all optimization solutions can meet the constraint conditions of vehicle collision response and structural impact characteristics.

[0105] After obtaining the multi-objective Pareto solution set, the entropy value weighting method is used to calculate the optimal feasible solution, and an orbital vehicle crashworthiness design scheme with controllable vehicle collision posture and stable energy absorption structure deformation is obtained. The comparison of various indicators before and after optimization is shown in Table 2.

[0106]

[0107] The above two results show that the optimization results obtained by using the calculation scheme have stable energy absorption structure deformation mode, controllable load fluctuation, and vehicle collision response indicators meeting the standard requirements, and are obviously improved compared with before optimization. The results fully demonstrate the effectiveness and reliability of the calculation scheme, and effectively realize the collaborative improvement of vehicle and structural crashworthiness.

[0108] The embodiment of the present application also provides an orbital vehicle and vehicle energy absorption structure crashworthiness collaborative optimization system, comprising a processor, a memory;

[0109] The memory is used to store a computer program;

[0110] The processor is used to execute the program stored on the memory, and realizes the method steps of any one of the orbital vehicle and vehicle energy absorption structure crashworthiness collaborative optimization method.

[0111] The orbital vehicle and vehicle energy absorption structure crashworthiness collaborative optimization system described above can realize each embodiment of the orbital vehicle and vehicle energy absorption structure crashworthiness collaborative optimization method described above, and can achieve the same beneficial effects, which will not be described here.

[0112] The above detailed the preferred embodiments of the present application. It should be understood that those skilled in the art can make many modifications and changes without creative labor based on the concept of the present application. Therefore, any technical solution obtained by logical analysis, reasoning or limited experiment based on the existing technology according to the concept of the present application shall be within the protection scope determined by the claims.

Claims

1. A method for synergistic optimization of the crashworthiness of rail vehicles and their energy-absorbing structures, characterized in that, include: Impact tests and finite element simulations of energy-absorbing structures were conducted to obtain a dataset of impact characteristics of energy-absorbing structures under different structural parameter configurations. Based on the dataset of impact characteristics of energy-absorbing structures, a structural hierarchical deep proxy model was established to predict the load characteristic curves and deformation mode images of energy-absorbing structures. The load characteristic curve of the energy-absorbing structure output by the structural hierarchy deep proxy model is input into the vehicle collision dynamics model to carry out vehicle-level collision simulation, and a vehicle hierarchy deep proxy model is constructed to predict the vehicle collision response. Based on the deformation pattern image obtained from the structural hierarchy deep proxy model and the vehicle collision response obtained from the vehicle hierarchy deep proxy model, a multi-objective rail vehicle / energy-absorbing structure collision resistance collaborative constraint optimization is performed to obtain a rail vehicle collision resistance optimization scheme with stable deformation of the energy-absorbing structure and controllable vehicle attitude evolution. The steps for establishing a hierarchical deep proxy model include: Based on the configuration type of the energy-absorbing structure of the rail vehicle, the key parameters affecting the collision energy absorption response of the structure are determined, structural impact tests and corresponding numerical simulations are carried out under different parameter configurations, and the simulation model is calibrated. Based on the numerical simulation model after experimental benchmarking, an impact characteristic dataset of energy-absorbing structures was generated and divided into training set, validation set and test set; Based on a deep learning framework and a dataset of energy-absorbing structure impact characteristics, a structural hierarchical deep proxy model is built. The training set is used to train the structural hierarchical deep proxy model, and the validation set and test set are used to validate and test the trained structural hierarchical deep proxy model. The structural hierarchical deep proxy model includes an impact load curve network and a deformation mode image network. The impact load curve network is used to predict the load characteristic curve of the energy-absorbing structure, and the deformation mode image network is used to predict the deformation mode image. The construction of the impact load curve network includes: Construct an autoencoder network with stacked long short-term memory layers as the main architecture, and extract the latent variables from the load characteristic curve; A multilayer perceptron vector prediction network is constructed to predict the latent variables of the corresponding extracted load characteristic curves from the structural parameters, and the load curves are output by decoding the decoder in the autoencoder network. The construction of the deformable mode image network includes: An autoencoder network with stacked convolutional layers as its main architecture is constructed to extract latent variables from deformable pattern images; A multilayer perceptron vector prediction network is constructed to predict the latent variables of the extracted deformable pattern image based on the structural parameters, and then decoded by the decoder in the autoencoder network to output the deformable pattern image.

2. The method for collaborative optimization of the crashworthiness of rail vehicles and vehicle energy-absorbing structures according to claim 1, characterized in that, The steps for constructing the vehicle-level deep proxy model include: The load characteristic curve of the energy-absorbing structure output by the structural hierarchy deep proxy model is input into the vehicle collision dynamics model. Vehicle impact tests and corresponding numerical simulations are carried out under different parameter configurations to calibrate the simulation model. Based on the numerical simulation model after experimental benchmarking, a vehicle collision response dataset is generated and divided into training set, validation set and test set; Based on a deep learning framework and a constructed vehicle collision response dataset, a vehicle-level deep proxy model is built, using energy-absorbing structure parameters as input to predict vehicle collision responses.

3. The method for collaborative optimization of the crashworthiness of rail vehicles and vehicle energy-absorbing structures according to claim 1, characterized in that, The steps of the multi-objective rail vehicle / energy-absorbing structure crashworthiness collaborative constraint optimization include: The design variables are the key parameters of the energy-absorbing structure, and the optimization objectives are the maximum average deceleration and total energy absorption during the vehicle collision process. The constraints include constraints on the impact characteristics of the energy-absorbing structure and constraints on the vehicle collision response. The constraints on the impact characteristics of the energy-absorbing structure include constraints on the stability of the deformation mode and constraints on the load stability. The constraints on the vehicle collision response include constraints on the maximum vertical lift of the wheelset. Multi-objective constrained optimization is performed using non-dominated sorting genetic algorithm, particle swarm optimization algorithm or differential evolution algorithm to obtain Pareto solution set, and the optimal feasible solution is calculated using shortest distance method or entropy weighting method.

4. The method for collaborative optimization of the crashworthiness of rail vehicles and vehicle energy-absorbing structures according to claim 3, characterized in that, The deformation mode stability constraint in the impact characteristic constraint of the energy-absorbing structure is defined as follows: If the deformation mode is stable, return constraint value K=0; if the deformation mode is unstable, return constraint value K=1.

5. The method for synergistic optimization of the crashworthiness of rail vehicles and vehicle energy-absorbing structures according to claim 3, characterized in that, The load stability constraint in the impact characteristic constraint of the energy-absorbing structure is defined as follows: ； In the formula, S 2 It serves as an evaluation index for the volatility of the load characteristic curve. θ This represents the load stability constraint value for the energy-absorbing structure. N This represents the number of sampling points for the load curve. F i For the load value at any sampling time, i Sampling time, This represents the average value of the load curve; If S²≤θ, return constraint value 0; if S²>θ, return constraint value 1, where θ is the preset constraint value.

6. The method for synergistic optimization of the crashworthiness of rail vehicles and vehicle energy-absorbing structures according to claim 3, characterized in that, The multi-objective constrained optimization using non-dominated sorting genetic algorithm, particle swarm optimization algorithm, or differential evolution algorithm to obtain the Pareto solution set includes: Based on the design variables, optimization objectives, and constraints, a mathematical model for the collaborative optimization of the crashworthiness of the rail vehicle / energy-absorbing structure is constructed. The mathematical model satisfies the following relationship: ； In the formula, X represents the energy-absorbing structural parameter, X L and X U These are its upper and lower boundaries, respectively. θ , where K is the load stability constraint value of the energy-absorbing structure, K is the mode stability constraint of the energy-absorbing structure, and MVL is the vehicle wheelset lift constraint; The constructed mathematical model for the collaborative optimization of the crashworthiness of rail vehicles / energy-absorbing structures is solved using a non-dominated sorting genetic algorithm, particle swarm optimization algorithm, or differential evolution algorithm to obtain the Pareto solution set; The maximum wheelset vertical lift constraint in the vehicle collision response constraints is defined as follows: According to the crashworthiness design standards for rail vehicles, the maximum vertical lift of the wheelset during a collision should be ≤21mm.

7. A system for collaborative optimization of the crashworthiness of rail vehicles and their energy-absorbing structures, characterized in that, Including processor and memory; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the steps of the method described in any one of claims 1-6.

Images